Abstract: Aiming at the efficiency of knowledge fusion, the authors extended the structure of conventionaltopicmaps. This new structure embodied the multilevel, multi-granularity and inherent relevant characteristics of knowledge. A new distributed knowledge fusion architecture based on extended topic maps was built up, and a novel algorithm of the similarity of topic maps based on comprehensive information was proposed by integrating syntactic similarity, semantic similarity and pragmatic similarity. It gave full consideration to the meaning and context of the compared elements so as to improve the accuracy of the similarity algorithm. The fusion rules and algorithm for extended topic maps were presented, implementing the effective knowledge fusion in distributed environment.

Key words: knowledge fusion, topic maps, knowledge element, similarity algorithm